From: Bennett Todd (no email)
Date: Sat Jan 01 2005 - 08:37:57 EST
2005-01-01T10:55:23 Craig Sanders:
> probably the three most important things missing are no tcp:53,
> [...]
You must be repeating someone else's uninformed criticism. djbdns of
course supports tcp, always has, that's a basic requirement for
either of the dns protocols.
> [...] no axfr, [...]
always had that, too, client and server.
> [...] and no ability to parse or use bind zonefiles.
That, I think, can be forgiven, considering what a lousy,
error-prone format it is, how much code is consumed trying to parse
it, and that each major implementation parses it differently.
When you need to work with BIND format zone data, find out which
version of BIND it's known to work with, use a temporary run of
that version of BIND to parse it and publish it for axfr, then zone
transfer it to tinydns-data's format with axfr-get.
> > djbdns implements a useful and safe subset, correctly.
>
> except for the bits that he doesnt like....which especially
> includes anything that facilitates inter-operation with non-djb
> software.
Bizarre. I've got to assume you haven't tried to use it or read
anything about it, to make statements like this. Almost no install
of DNS would be useful in any way if it weren't perfectly
inter-operable with other implementations.
Yes, there are parts of the DNS protocol that djbdns doesn't
implement. The same can be said of every implementation in
existence; the part that people actually use is a tiny subset of the
whole, and people are continuing to devise more and more unused bits
of DNS protocol every day (see, for example, the DNS security
extensions that were designed without working key management).
But the parts that people need to publish or resolve DNS data all
work fine, and are nice and simple and efficient and robust. Used by
lots of major sites, all over the place.
-Bennett
|
|
|